Color and Texture Information Processing to Improve Storage Beans
Ngatchou Alban *
Institute of Agricultural Research for Development (IRAD), P. O. Box 2067 or 2123 Yaoundé, Cameroon and Modelisation, Image Processing and Applications Research Group (MOTRIMA), Biophysics and Food Biochemistry Laboratory, National School of Agro-Industrial Sciences, University of Ngaoundéré, P. O. Box 455, Ngaoundéré, Cameroon
Bitjoka Laurent
Modelisation, Image Processing and Applications Research Group (MOTRIMA), Biophysics and Food Biochemistry Laboratory, National School of Agro-Industrial Sciences, University of Ngaoundéré, P. O. Box 455, Ngaoundéré, Cameroon
Boukar Ousman
Modelisation, Image Processing and Applications Research Group (MOTRIMA), Biophysics and Food Biochemistry Laboratory, National School of Agro-Industrial Sciences, University of Ngaoundéré, P. O. Box 455, Ngaoundéré, Cameroon
Tonye Emmanuel
National School of Engineering, Electronics and Signal Processing Laboratory (LETS), University of Yaoundé 1, P. O. Box 8390, Yaoundé, Cameroon
*Author to whom correspondence should be addressed.
Abstract
Aims: This paper attempts to improve automatic temporal change detection on a pair of beans images, acquired before and after storage under high temperature (≥ 25°C) and high relative humidity (≥ 65%), conditions that promote « Hard-To-Cook » phenomenon.
Study Design: Image processing, Hard-To-Cook beans.
Place and Duration of Study: Laboratory of Modelisation, Image Processing and Applications Research (MOTRIMA) Department of Electrical Engineering Energetic and Automatics, Laboratory of Biophysics and Food Biochemistry Department of Food Science and Nutrition of National School of Agro-Industrial Sciences (University of Ngaoundéré, Cameroon), Institute of Agricultural Research for Development (IRAD) between August 2009 and March 2010.
Methodology: We want to get a robust extracting seed in acquired images and a good dissimilarity parameter for temporal change detection on a pair of textured images. To reach this goal, we analyze the characterization of textural properties and space color which are more relevant to textured beans seeds. We use wavelet transform and apply fuzzy logic segmentation. We define a confidence limit for the dissimilarity parameter before analyzing its evolution during storage of beans seeds. Finally we correlate this parameter with another Hard-To-Cook indicator.
Results: After many tests, Daubechies 2(db2) wavelet family in RGB space allowed best extracting beans seeds in scene with fuzzy-c-means segmentation. The global intensity variation was a pertinent parameter for dissimilarity detection between two images. We obtained highly correlation between this parameter and cooking times beans (-0.96; -0.88; -0.72 respectively in Red, Green and Blue color space).
Conclusion: The global intensity variation in red color space allowed the determination level of browning beans seeds as indicator of their Hard-To-Cook degree.
Keywords: Texture and color image processing, wavelet transform, global intensity variation, ECA PAN 019 beans (Phaseolus vulgaris), hard-to-cook, fuzzy logic image segmentation